Luis Fos

3D Designer / Developer. Mostly Houdini

[tinybugbot @tinybugbot]

modeling 🐈
#Houdini

[Luis Fos @luisfoscg]

@_nscr Is it using attribute AOV to drive COP streak blur?

[Google DeepMind @GoogleDeepMind]

What if you could not only watch a generated video, but explore it too? 🌐

Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt.

From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵

[Pasquale Minervini @PMinervini]

"in 2025 we will have flying cars" 😂😂😂

[kyutai @kyutai_labs]

Kyutai TTS and Unmute are now open source!
The text-to-speech is natural, customizable, and fast: it can serve 32 users with a 350ms latency on a single L40S. Try it out and get started on the project page: https://t.co/B4P9FuOrQc

[Luis Fos @luisfoscg]

@KMkota0 Which model do you use, Demucs?

[RIO @riomadeit]

cant believe humans saw this then created spotify

[Luis Fos @luisfoscg]

@henningsanden These kind of quick tips are really valuable

[Luis Fos @luisfoscg]

@dgant I agree, it’s easy to spread focus too thin.
But on the upside, unfinished projects can often teach you a lot, and ideally that momentum carries into the next goal.
Nothing is ever “finished” anyway

[Luis Fos @luisfoscg]

@BockJakob would love to play around with this 👀

[Luis Fos @luisfoscg]

Mental to see artists at a company this big rip off another artist so openly.
What were they thinking?

[N² @4nt1r34l]

the Marathon alpha released recently and its environments are covered with assets lifted from poster designs i made in 2017.

@Bungie @josephacross

[Luis Fos @luisfoscg]

@uns__nstudio very cool🔥though it freezes often for me

[Luis Fos @luisfoscg]

@henningsanden There is the CAI C2PA push to store metadata, don't know how it helps an end user though.

Reverse image checkers are probably our best bet for now

[Luis Fos @luisfoscg]

@billy_xtz @sidefx The shape is static, so calculate UVs on final frame then transfer to the deforming mesh

[Luis Fos @luisfoscg]

Working on some interactive visuals for my portfolio site.

Using Three.js + WebGL to pass mouse position and velocity into a ping-pong buffer for image distortion. Quite a lot of gotchas to setup with GLSL 3.0.

Any ideas for interactions?

[Luis Fos @luisfoscg]

@henningsanden I wonder how we can progress to simple communication. Like we’d need to first get so dependent on AI summaries, then suddenly lose all access to AI but remain speaking in prompts

[Luis Fos @luisfoscg]

@RohanCreates Is the lighting done manually through cops per source - or can you pull a direct lighting aov?

[Luis Fos @luisfoscg]

@TimSweeneyEpic Nice to see a major player actually pushing for pro-competition, pro-consumer change

[Luis Fos @luisfoscg]

@henningsanden 100 people go to the airport.
Each one takes a cab, 100 cars.
Now some share cabs together, 50 cars.
Put everyone on a big bus, 1 vehicle.

The goal is get everyone to the airport in as few cars as possible, despite their varying flight times.

People = data, Cars = draw calls

[Luis Fos @luisfoscg]

@henningsanden @LoganCDN Reduce polycount, tri mesh, avoid slivers.
Abuse instancing: reuse meshes, uvs, materials when possible.
Maximise UV island space, minimise texture resolution.
My view is send as little data to GPU as possible, though I’m not as experienced in games.